3

In my application I have the signals from 6 analog MEMS microphones (CUI Devices CMM-3729AT-42108-TR), each amplified by a pair of op-amps in a voltage follower -> inverting amp configuration. The microphones are on tiny PCBs buried deep within a chemical analysis cavity, and the only way to get the signals out is through a short cable (150mm, conductors are +3V3, SIGNAL, and GND). After amplification, it's off to an ADC to be digitized, and then I'm demodulating the signal to sniff out the amplitude of a ~1500 Hz signal.

Microphone Amplification circuit

Due to the stupidly small space I have to deal with (can't change the geometry of the analysis cavity), I'm prototyping with the Molex PicoBlade system. DigiKey sells pre-fabricated cable assemblies, but they're all unshielded. Many of the cables are touching each other.

Should I be considering a way to shield the cables? The DC blocking capacitor is near the op-amp end of things, so the signal on the SIGNAL wire will be biased around +550 mV. This can be changed, I can move the DC blocking cap closer to the microphone if recommended. If I should be shielding, what're some recommendations for that?

Did I miss anything else in the circuit? The microphones each have power supply bypass caps very close to them (though not shown in the circuit I've included). The output impedance of the microphones is 300 ohms at 1 kHz. How does that interact with the input impedance of the op-amp?

Ben S.
  • 335
  • 1
  • 3
  • 12
Ben S.
  • 71
  • 3

5 Answers5

1

My opinion only: so long as the output impedance of the preamp is fairly low, you should NOT need to provide shielding between the output of the preamp on the card containing the microphones and the other card containing the ADC.

Several factors affect noise pickup.

First is the actual level involved. The higher the desired level sent down the conductor, the less effect external noise has. It's a matter of ratios.

2) Output impedance of the source determines just how much external noise will affect the signal. In an ideal world, zero output impedance is immune to external EMI fields. We don't live in an ideal world but in general, the lower the source impedance, the less effect external EMI has on the signal.

Dwayne Reid
  • 23,390
  • 2
  • 35
  • 67
  • 1
    There isn't a preamp on the microphone card. It's a small PCB, barely large enough to hold the microphone and the plug for the 3-pin connector. The output impedance of the mic itself is 300 ohms. The signals up the cable are expected to be in the 0-2 mV range. – Ben S. Feb 14 '20 at 22:14
  • 1
    Gotcha. You say the output impedance of the sensors is about 300 Ohms. Still most likely not a problem. – Dwayne Reid Feb 15 '20 at 02:47
  • 1
    even if some of the cables from different microphones are touching each other? – Ben S. Feb 15 '20 at 03:00
  • 1
    Still should not be a problem. – Dwayne Reid Feb 15 '20 at 03:01
1

As others say, shielding might not be necessary... But watch out for voltage drop on the ground wires, if the current in them changes. For example if something else on your board turns on, that current step will transfer to all microphone outputs.

Large power supply capacitors on the board will help. Separating power ground and signal ground, and using differential amplifiers before the ADC will solve this.

Check out the U.Fl RF connector - it's much smaller than a picoblade and is well shielded. It doesn't solve the common current problem though.

tomnexus
  • 7,617
  • 3
  • 19
  • 35
  • Those U.FI connectors look perfect for small spaces, but I need 3 conductors! I need to get power down to the microphones as well. – Ben S. Feb 16 '20 at 20:24
  • 1
    @bens oh I only meant for the signal, the power would still need a cable. It might be possible to send the power down the centre of the coax, pick it off with an inductor. Or use two coaxes? Look at all the coax cables in [this quantum computer](https://www.cnet.com/pictures/take-a-look-at-googles-quantum-computing-technology/)! – tomnexus Feb 16 '20 at 21:37
  • 1
    @bens Have you considered long thin flex PCBs? 150 mm x 3 mm or so. With two layers you would get good ground and good shielding between boards if they were stacked the right way. Use board-to-board connectors on the top end to connect to the amplifier/ADC board? – tomnexus Feb 16 '20 at 21:45
0

The unshielded short cable is 300 ohms so it probably will not pickup much interference. The voltage follower opamp does nothing and should be removed. The input resistor of the inverting opamp can use the 1.6k resistor then the 0.1uF capacitor feeding it cuts frequencies below 1kHz.

The 2.2uF output capacitor will cause the inverting opamp to oscillate. To cut high frequencies a capacitor can be parallel with the negative feedback resistor.

Audioguru
  • 3,894
  • 5
  • 8
  • 1
    I added the voltage follower so I can use smaller values of R_INPUT and R_FEEDBACK in the gain stage - it's a recommended addition for MEMS circuits (Analog Devices AN-1165, figure 7). Thanks for the note about a cap in parallel with R_FEEDBACK, I'd meant to include it, but I didn't see any oscillations in Spice simulations. I noted in the OP that this is a fixed-frequency application of ~1500 Hz. If I made the feedback circuit into an RC low-pass filter with ~2500 Hz cutoff, would that be sufficient? – Ben S. Feb 14 '20 at 22:12
0

Your input highpass filter and feedback lowpass filter have simple gradual slopes so they actually reduce the level of your 1500Hz signal and also do not make a good bandpass filter. A multiple-order Sallen-Key highpass and lowpass filters make a good bandpass filter. A Wien Bridge, Multiple Feedback or Twin-T bandpass filter has a high Q at its peak with sharp cutoffs but reduces its slopes when away from the peak frequency.

Audioguru
  • 3,894
  • 5
  • 8
0

You don't say what level of control you have over the design, but you might consider MEMS mics with a digital (PDM) output. Many audio ADCs will take this directly; if not, you can easily resample and LPF the signal to recover the analog version.

If you have particular noise concerns (crosstalk, motor noise, line noise, etc.) edit them into your question and we could probably address them specifically.

Cristobol Polychronopolis
  • 9,622
  • 9
  • 17